fix for accelerate env var for auto bf16, add new base image and expand torch_cuda_arch_list support (#1413)

.github/workflows/base.yml

@@ -16,17 +16,22 @@ jobs:
             cuda_version: 11.8.0
             python_version: "3.10"
             pytorch: 2.1.2
-            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 9.0+PTX"
+            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
           - cuda: "121"
             cuda_version: 12.1.0
             python_version: "3.10"
             pytorch: 2.1.2
-            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 9.0+PTX"
+            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
           - cuda: "121"
             cuda_version: 12.1.0
             python_version: "3.11"
             pytorch: 2.1.2
-            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 9.0+PTX"
+            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
+          - cuda: "121"
+            cuda_version: 12.1.0
+            python_version: "3.11"
+            pytorch: 2.2.1
+            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
     steps:
       - name: Checkout
         uses: actions/checkout@v3

src/axolotl/utils/trainer.py

@@ -11,6 +11,7 @@ import torch.cuda
 from accelerate.logging import get_logger
 from datasets import set_caching_enabled
 from torch.utils.data import DataLoader, RandomSampler
+from transformers.utils import is_torch_bf16_gpu_available
 
 from axolotl.core.trainer_builder import HFCausalTrainerBuilder, HFDPOTrainerBuilder
 from axolotl.utils.distributed import is_main_process, reduce_and_broadcast, zero_first
@@ -324,6 +325,11 @@ def prepare_optim_env(cfg):
         os.environ["ACCELERATE_USE_DEEPSPEED"] = "true"
         os.environ["ACCELERATE_DEEPSPEED_CONFIG_FILE"] = cfg.deepspeed
 
+    if (cfg.bf16 == "auto" and is_torch_bf16_gpu_available()) or cfg.bf16 is True:
+        os.environ["ACCELERATE_MIXED_PRECISION"] = "bf16"
+    elif cfg.fp16:
+        os.environ["ACCELERATE_MIXED_PRECISION"] = "fp16"
+
 
 def setup_trainer(cfg, train_dataset, eval_dataset, model, tokenizer, total_num_steps):
     if cfg.rl in ["dpo", "ipo", "kto_pair"]: